UNCLASSIFIED

RANDOM NOISE MONOPULSE RADAR SYSTEM FOR COVERT TRACKING OF TARGETS

Ram M. Narayanan
Department of Electrical Engineering
University of Nebraska
Lincoln, NE 68588-0511
Tel: (402) 472-5141; Fax: (402) 472-4732; email: ram@doppler.unl.edu

ABSTRACT of targets is ultra-wideband (UWB) random noise

The University of Nebraska is currently developing
a unique monopulse radar concept based on the use of
random noise signal for covert tracking applications.
This project is funded by the Missile Defense Agency
(MDA). The advantage of this system over
conventional frequency-modulated continuous wave
(FMCW) or short pulse systems is its covertness
resulting from the random waveform’s immunity from
interception and jamming. The system integrates a
novel heterodyne correlation receiver with conventional
monopulse architecture. Based on the previous work
such as random noise interferometry, a series of
theoretical analysis and simulations were conducted to
examine the potential performance of this monopulse
system. Furthermore, a prototype system is under
development to exploit practical design aspects of
phase comparison angle measurement. It is revealed
that random noise monopulse radar can provide the
same function as traditional monopulse radar, i.e.,
implement range and angular estimation and tracking in
real time. The bandwidth of random noise signal can be
optimized to achieve the best range resolution as well
as the angular accuracy.
INTRODUCTION
Phase comparison monopulse uses two apertures
with displaced phase centers to locate the angle of
arrival from scatters. The characteristic of this
technique is its dependence on the phase information of
received signals. When a random noise transmit
waveform is employed, there will be much higher phase
uncertainties compared to traditional waveforms due to
its random nature. A phase coherent processing
technique using the heterodyne correlation architecture
has been developed and applied towards polarimetry1,
Doppler estimation2, synthetic aperture (SAR) radar3,
and inverse SAR (ISAR)4 with good success. The
results obtained compare well with those obtained using
conventional waveforms with the added advantage of
covertness, i.e., immunity from detection and jamming.
One application we have demonstrated that clearly
suggests the use of this technique for angular tracking
interferometry. In our experiments, we showed that it
was indeed possible to use the phase difference between
spaced receiver antennas to locate a target in azimuth,
while precise range information was obtained from the
target delay. We have also developed the necessary
analytical formulation for a clearer understanding of
this technique together with its advantages and
limitations5.
Our recent results analyze the applicability of
conventional phase-comparison monopulse techniques
to the random noise radar system. A monopulse
architecture based on sum-and-difference network was
used to perform simulation studies. The transmit
waveform was assumed to be bandlimited white noise,
which was approximated as the summation of a large
number of frequency components over the bandwidth,
each component having a random amplitude. Received
signals are passed through the sum-and-difference
hybrid and mixed with a delayed replica of the transmit
signal. The intermediate frequency (IF) outputs are
routed through band pass filters following which a
complex correlation operation is performed. This output
provides information on the target direction
dynamically. A detailed analysis shows that under ideal
conditions, i.e., flat frequency characteristics for the
atmospheric propagation as well as for the target
reflectance over the operating bandwidth, the output of
the monopulse system is identical to that of the single
frequency monopulse in the average sense.
ANTENNA SYSTEM
The antenna is the first component to process the
received random noise signal plus uncorrelated system
noise. Two special factors influence the performance of
random noise monopulse antenna system. The first is
signal bandwidth, and the second is the random
fluctuations in signal phase and amplitude.
For narrowband systems, the antenna pattern is
generally well characterized, and is considered
invariant over the operating frequency range. However,
a UWB waveform operates over a much wider
fractional bandwidth, typically greater than 25%. The

1
Approved for Public Release;
Distribution is Unlimited.
UNCLASSIFIED
 Report Documentation Page
Report Date Report Type Dates Covered (from... to)
29JUL2002 N/A -

Title and Subtitle Contract Number

Random Noise Monopulse Radar System for Covert
Tracking of Targets Grant Number

Program Element Number

Author(s) Project Number

Task Number

Work Unit Number

Performing Organization Name(s) and Address(es) Performing Organization Report Number

Department of Electrical Engineering University of
Nebraska Lincoln, NE 68588-0511

Sponsoring/Monitoring Agency Name(s) and Sponsor/Monitor’s Acronym(s)

Address(es)
Sponsor/Monitor’s Report Number(s)

Distribution/Availability Statement
Approved for public release, distribution unlimited

Supplementary Notes
See Also ADM201460. Papers from Unclassified Proceedings from the 11th Annual AAIA/MDA
Technology Conference held in Monterey, CA from 29 Jul - 2 Aug 2002., The original document contains
color images.

Abstract

Subject Terms

Report Classification Classification of this page

unclassified unclassified

Classification of Abstract Limitation of Abstract

unclassified SAR

Number of Pages
9
 UNCLASSIFIED
differences in the antenna patterns at different
frequencies may introduce measurement errors. As an
example, Figure 1 shows the amplitude pattern of a
typical X-band horn antenna at 9 GHz, 10 GHz and 11
GHz. This simulation suggests that over a narrow
angular region (±10°), the antenna patterns over the
frequency range are essentially identical. However, the
patterns are different beyond this region.
(a) E-plane pattern
(b) H-plane pattern
Figure 1: Magnitude of pattern of an X-band horn
antenna at different frequencies (9, 10, and 11 GHz).
The above antenna patterns are still approximations
of the real behavior of a given antenna when used with
random noise waveforms. Due to the fact that different
frequency components occur in random manner, the
actual antenna response will vary with time. An
approach to solving this problem is through the use of
statistical antenna theory to determine the mean antenna
characteristics. Wideband excitation of antenna system
impacts the accuracy of the angular measurement by
imposing additional fluctuations in the sum and the
difference outputs. However, the effects of these
UNCLASSIFIED
variations can be reduced by increasing the observation
time and performing signal averaging.
BASIC MONOPULSE MODEL
Figure 2 shows the conceptual model of our
proposed random noise phase comparison monopulse
system architecture. The two receive antennas with
phase centers displaced by 2d feed their signals into a
sum and difference network. Unlike the traditional
monopulse receiver that mixes two channel signals with
a single frequency local oscillator, this architecture
correlates the sum and difference channel signals with a
delayed replica of transmitted random noise signal
X (t ) which we denote as s d (t ) .
sd (t )
s1 (t ) s Σ (t ) BPF
sΣI (t )
2d
+
( )
s ΣI s ∆I
* u out
− s ΣI
2
s2 (t ) s ∆ (t ) s ∆I (t )
BPF
sd (t )
Figure 2: Conceptual model for random noise monopulse
system.
In practice, since the delayed transmitted replica is
always downconverted to an intermediate frequency
(IF) before correlation, phase information is extracted at
the IF stage. For simplicity, we can disregard this
downconversion process and perform angle estimation
directly at baseband. Thus, the bandpass filters in
Figure 2 can be temporarily suppressed.
We assume that the target being tracked is far away
from antenna system at a range r and off-axis angle θ .
The transmitted band-limited white noise signal
X (t ) can be approximated as the sum of a large number
of frequency components across a wide bandwidth, i.e.,
M
X̂ (t ) = ∑A e
i =1
i
jω i t
(1)
In (1), M can be an arbitrary large number
depending on the choice of interval ω 0 between
 UNCLASSIFIED

adjacent frequency indices, while A i denotes the simplification, the time average of monopulse output
random variable satisfying the property of non-zero can be expressed as
autocorrelation and zero cross-correlation. Ignoring
propagation effects on amplitude of the different M ω ω
j i dsinθ M − j i dsinθ
frequency components, the received signal at antenna
with coordinate value d k (k = 1,2) can be described as
∑ 2⋅
Ai e c − 2
∑
Ai ⋅ e c
u out = i =1 i =1 (6)
ωi M ωi ω
M −j (2 r + d k sinθ ) j dsinθ M − j i dsinθ
X k (t ) = ∑ A i e jω t ei c (2) ∑ 2⋅
Ai e c + 2
∑
Ai ⋅ e c
i =1 i =1 i =1

where c is the velocity of light. Here, A is the mean value of random amplitude A .
i i
Ignoring some amplitude coefficients introduced by Recall that the transmit signal is assumed to have
the hybrid components, the outputs of sum and constant power density across the total bandwidth ∆ω .
difference network are given by In other words, for each frequency component ω i , we
have
s Σ (t ) = X1 (t ) + X 2 (t ) ,
(3)
s ∆ (t ) = X1 (t ) − X 2 (t ) 

ω 0S 0
m=n
A i2 =  2π (7)
The above sum and difference signals are mixed  0 m≠n
with the delayed replica s d (t ) . The correlation results
in both channels are routed through filters that retain Thus, (6) can be simplified by taking summations
the low frequency components around the IF. These term by term. Assuming that all the frequency
filtered outputs are then used to form the complex components are equally distributed around the center
correlation coefficient in order to generate the final frequency ω c , and that ω 0 is very small, the final
angle estimation. Consequently, the final output of expression derived using (6) and (7) is
phase comparison monopulse at any given time is
obtained as ω 
u out ≈ tan  c dsinθ  (8)
 c 
 
 sI sI * 
u out = Im Σ ∆ 
( ) (4) We see that (8) has the same form as the
 I 2  characteristic curve of a phase comparison monopulse
 sΣ  system using a single frequency signal. This means that
 
we can use the random noise waveform to achieve the
where Im(.) denotes the imaginary part of the complex same angle measurement as conventional monopulse by
taking time average, as long as the delay line is ideal.
argument. Note that s IΣ and s ∆I
are the low frequency
outputs in the sum and difference channels respectively. In actual practice, a real delay line always has finite
range resolution, while the moving target will introduce
We assume that the delay line is able to delay the phase errors. These factors decrease system sensitivity,
transmitted replica by the precise time according to dynamic range and tracking accuracy. For this case, let
range 2r, i.e., us consider a “bad delay line”, which does not provide
any delay and also loses all the frequency components.
ωi This is equivalent to saying
M −j ⋅2 r
sd (t ) = ∑ Aie jω i t
e c (5)
s IΣ (t ) = ρs Σ (t ) ,
i =1
(9)
I
s∆ (t ) = ρs ∆ (t )
Then, following the correlation and filtering operations,
the range-dependent phase is eliminated, and only the
phase containing the angle information remains. Based Given (2), (3) and (9), we recomputed the
on the properties of coefficients in (1) and (4), and after monopulse output as shown in (4). This yields

UNCLASSIFIED
 UNCLASSIFIED

M ω 
ω  sin c 2 dsinθ 
∑ A i2 sin  i ⋅ 2 dsinθ 
 c  u out =  c  = tan  ω c dsinθ 
 
u ′out = i =1  ωc 
M M
(10)
1 + cos 2 dsinθ   c 
ω  (16)
∑ Ai2 + ∑ A i2 cos i ⋅ 2 dsinθ 
 c 
 c
 2π


i =1 i =1
= tan  dsinθ 
 λc 
Without loss of generality, we can let M, which is
large, be an odd number, i.e., we can assume M=2N+1. Again, this is the same as the single frequency
This means there are N frequency components below monopulse output.
and N frequency components above the center
frequency ω c . Keeping (7) in mind, we can transform We can also note that when the signal has narrow
(10) into bandwidth, i.e., ∆ω → 0 , (15) and (16) is applicable
ω  naturally. Thus we revert to the case of single frequency
sin  c 2 dsinθ  monopulse.
u out =  c  (11)
 ωc 
K c + cos 2 dsinθ  In summary, even for the worst case, the angle of
 c  arrival can still be estimated using the general
where expression (11), although the receiver sensitivity is
lowered. For UWB signals, it is also possible to obtain
the same tracking capability as the single frequency
M monopulse by choosing the appropriate bandwidth.
Kc =
sin[Nω 0 dsinθ c]
1+ ⋅ 2 cos[(N + 1)ω 0 dsinθ c]
sin[ω 0 dsinθ c]
SYSTEM DESCRIPTION
(12) A complete system has been designed to operate at
X-band for demonstrating the random noise phase
As the frequency separation ω 0 approaches zero, comparison concept. The radar system adopts a phase
the above equation more accurately represents the coherent heterodyne architecture, and uses two parallel
random noise monopulse system. Furthermore, channels to perform single plane tracking. Figure 3
shows the simplified block diagram of this radar.
M
Kc = (13)
 ∆ω 
1 + 2 N⋅ cos dsinθ 
 2c  BPF
100~200Mhz
BPF
10.15~10.25G
Power AMP
Horn
Antenna
Noise
Source
Phase-locked
It can be seen from (11) and (13) that the Frequency
Source 70M AMP

monopulse output depends on the signal bandwidth.

Generally K c > 1 and the monopulse characteristic
AMP
Delay&AMP
Phase Locked Dielectric
3dB
Resonator Oscillator
Attenuator
curve is flattened due to this term. However, if we
choose the bandwidth that precisely satisfies BPF
170~270M

c
∆ω = k⋅ 2 π sinθ k = 1,2,3,...
AMP
(14)
d
IF AGC
IF BPF Receiving Horn
the cosine term will goes to 1 and IQD1
AMP

4~12.4G
100~200M BPF 180

M Degree

Kc → =1
hybrid
(15) IQD2

1+ 2N

Then (11) reduces to

Figure 3: Radar system architecture.

UNCLASSIFIED
 UNCLASSIFIED
The noise source generates bandlimited random
noise signal over the 100-200 MHz frequency range
through a bandpass filter. In the transmit channel, this
signal is amplified and upconverted to X-band over the
10.15-10.25 GHz frequency range. Another portion of
random noise signal is delayed and modulated as the
correlation signal s d (t ) .
Two receive antennas are placed apart with
distance 2d = 30 cm. The signals received by these
antennas are fed into the sum and difference network,
which is implemented by a 180° hybrid. Signals in the
sum and difference channels are first downconverted to
the 100-200 MHz frequency range, amplified, filtered,
and finally correlated with s d (t ) .
The final processing stage consists of extracting the
amplitude and phase information in the sum and
difference channels. This is implemented using two I/Q
detectors. The following signal/data processor digitizes
the I/Q signals and performs real-time closed-loop
control.
It is interesting to note that the IF bandpass filter
actually functions as a time averager that is essential for
random noise radar processing. Narrower the passband,
longer is the averaging time span. However, some
useful information (such as Doppler frequency shift)
may be lost if we use too narrow a bandwidth, and a
filter with very narrow relative bandwidth is generally
difficult to implement. For these reasons, some time
averaging is still needed in digital processing stage.
There are some special requirements that make
random noise monopulse signal processor different
from other noise radars. One is the feedback control
capability, and the other is real time processing. The
processing scheme is described in Figure 4.
Data acquisition
Pre-processing
Complex correlation Angle extraction Control signal
&Normalization &Tracking generating
Figure 4: Data processing block diagram.
UNCLASSIFIED
A real time controller and a multifunction board are
adequate for the above tasks. Time constraints must be
considered to specify the required hardware/software
performance. For range delay and tracking functions,
the control cycle should be as short as possible since the
target moves continuously. This is a challenge not only
for the processor, but also for the delay line subsystem.
For angle tracking, the control cycle ∆T must satisfy
θ& ⋅ ∆T < θ M (17)
Here, θ& denotes angular velocity of the target, θ
M
is the linear region limitation of monopulse
characteristic curve. If the off-axis angle exceeds θ ,
M
the angular detector response will saturate, and the
antenna controller based on complex correlation will
not give correct control signals. More advanced data
processing design may use special algorithms to judge
the accuracy of the detected angle value, or smooth the
angle measurement using tracking filters.
SIMULATIONS AND OBSERVATIONS
Before the implementation of random noise
monopulse radar system, it is helpful to predict the
dynamic behavior of system architecture through
simulations. A rapid and accurate system simulation
raises many new requirements for simulation tools. As
a comprehensive dynamic system analysis environment,
®
SystemView is one such software that is suitable for
this application. A large system is first divided into
subsystems and then specified by tokens and
metasystems. Next, components are connected to each
other, with some analysis and visualization viewers
added. Once the simulation is started, it will go through
the system running process in real time.
Figure 5 depicts the random noise waveform
generated by a Gaussian noise source and a bandpass
filter with different filter bandwidths. Figure 5(a) shows
the Gaussian noise waveform generated by the noise
source and this simulates an infinite bandwidth. Figures
5(b) and 5(c) show the noise source output filtered
using filter bandwidths of 100 MHz and 1 GHz around
a center frequency of 10 GHz. These simulations verify
that the transmit signal bandwidth can be controlled
using the bandpass filter, and that a wider bandwidth
adds more fluctuations to the waveform envelope.
 UNCLASSIFIED
® implementation. Figure 6 is the top-level view of the
(a) Gaussian white noise simulated by SystemView
(b) Transmitted random noise waveform filtered using 100
MHz bandwidth BPF
(c) Transmitted random noise waveform filtered using 1 GHz
bandwidth BPF
Figure 5: Transmit waveform examples.
The transmitted signals will, in general, be
modified by the propagation media such as the
atmosphere. Basically, some specially designed filters
in one or two token modules can simulate this
phenomenon. Also, we can simulate additive
UNCLASSIFIED
interferences and jamming by some other sources and
adder modules. The combination of all above factors
and range-delay of return waves, received signals in
multiple channels can be simulated.
All the radar system components, including sum
and difference network (which can be implemented by
hybrids or 180° power combiners), amplifiers, mixers,
filters, power dividers, etc., are specified with different
performance and parameters, and inserted into
simulation system as modules. Thus, we can test a
broad range of system designs without actual physical
simulation system that includes several subsystems.
Figure 6: Top level simulation architecture for random
noise monopulse.
The architecture in Figure 6 is a “basic” configuration
in the sense that it is only used to simulate the
measurement of single plane angle associated with the
direction of the static target. With more complexity
added, it can be expanded to simulate real angle
tracking in dual planes.
Figure 7 (a) shows the subsystem architecture of the
sum and difference network, Figure 7(b) shows the
subsystem architecture of the I/Q detector, and Figure
7(c) shows the subsystem architecture of angle
extractor.
One of interesting points of the simulation
architecture is the choice of the intermediate frequency
(IF) and the bandpass filter following the correlation.
These two design factors influence the performance of
noise radar dramatically. In the next simulation, the
transmitted signal has a bandwidth of 1 GHz and the IF
is set at 300 MHz. We then observe the output of
correlator with various bandpass filter bandwidths, viz.,
 UNCLASSIFIED

2 MHz, 1 MHz, and 100 kHz. These are shown in deduce the impact of narrow bandpass filter on this
Figures 8(a), 8(b), and 8(c), respectively. process from Figure 8. We note that decreased

(a) Sum and difference subsystem

(a) Correlator BW=2 MHz

(b) I/Q detector subsystem

(b) Correlator BW=1 MHz

(c) Angle extractor subsystem

Figure 7: Subsystem simulation architectures.

In fact, when the random noise monopulse radar is (c) Correlator BW=100 kHz
used for range tracking, what we are interested in the
Figure 8: Real-time output waveform of correlator for
output of correlator is its envelope. We accomplish this
different correlator bandwidths.
by recording the envelope over time and computing its
average amplitude. The peak amplitude can be obtained bandwidth means more integration time; thus, we can
as long as the delay line is matched to the range delay obtain a smoother correlation output.
due to the two-way propagation to the target. We can

7
UNCLASSIFIED
 UNCLASSIFIED

However, we should be aware that too narrow a The effects of expanded signal bandwidth on receive
bandwidth for the filter could become the limiting antenna, receiver architecture, and signal processing are
factor for the system dynamic range. If the target studied. A prototype system is introduced and analyzed.
maneuvers abruptly, although the delay line response Before the actual implementation of system, a series of
may follow it in range, the narrowband filter response
may lag so much so as to lose the target in angle.
Therefore, an optimal system design should partition
the time averaging task appropriately between the
correlator and the subsequent digital processing system.

Now we turn to the results of angle measurement

output. It should be repeated at this point that due to the
random nature of transmitted signal, all the
measurements are in the sense of time average.
Generally, the final output of random noise radar will
fluctuate more severely than that of the single
frequency system.

Figure 9 compares the system output of single (a) System output of single frequency monopulse
frequency (10 GHz) monopulse and random noise
monopulse (IF BW=1 MHz). It is clear that for case of
single frequency system, phase information is extracted
quickly and output adjusts to a stable state with
rapidity, while the random noise monopulse output does
not attain very stable states. Rather, it fluctuates around
some mean value that is sensitive to off-axis angle of
the target.

In order to examine the angular characteristic curve

of coherent random noise monopulse, the following
simulations are performed for a spacing of 30 cm
between the two receive antennas. Figure 10(a) shows
the worst case system theoretical output as a function of
the transmit bandwidth at a center frequency of 1.5
GHz. The curves for a bandwidth of up to 500 MHz are
very similar to that of the single frequency (zero
bandwidth) case. In Figure 10(b), the angle of arrival
region between –0.25 and +0.25 radians is expanded:
the worst-case theoretical curve is shown in dotted
while the simulated curve is shown in bold. The
difference between these curves is attributed to the non-
ideal parameters assumed for the RF components.
Figure 10(c) shows curves similar to that in Figure
10(a), except the transmit frequency is now 10 GHz.
While the sensitivity is better, the dynamic range is
limited, as expected. Figure 10(d) shows the simulated
output at 10 GHz.
CONCLUSIONS
This paper introduces an innovative coherent
random noise radar system implementing phase
comparison monopulse. The theoretical analysis is
based on treating random noise waveform as random
summation of a great amount of frequency components.
(b) System output of coherent random noise monopulse
Figure 9: Comparison of system output of single
frequency monopulse and coherent random noise
monopulse.
simulations are performed, and the basic performance
measures of the random noise monopulse system are
outlined.
Generally speaking, the disadvantages of the random
noise monopulse radar are delayed response time,
lowered sensitivity, and restricted dynamic range
compared to single frequency or narrowband systems
operating at the same center frequency. However, they
can be overcome greatly by employing high accuracy
delay line and high speed time averaging. The system
has its main advantage in the fact that it possesses
electronic counter counter measures (ECCM)
capability, as has been clearly demonstrated by our
group6.

UNCLASSIFIED
 UNCLASSIFIED
(a) Theoretical output at 1.5 GHz frequency
(b) Same as above with expanded scale
Figure 10: Theoretical and simulated outputs of the random noise monopulse radar system.
ACKNOWLEDGMENTS
This project is supported by the MDA through the
U.S. Army Research Office under contract number
DAAD19-01-1-0531. We appreciate helpful comments
and suggestions from Dr. Juergen Pohlmann of MDA.
Thanks are also due to Yan Zhang and Xiaojian Xu for
assistance with the system development and simulation.
REFERENCES
1. R.M. Narayanan, Y. Xu, P.D. Hoffmeyer, and
J.O. Curtis, “Design, performance, and
applications of a coherent ultra-wideband
random noise radar,” Optical Engineering,
37(6): 1855-1869, June 1998.
2. R.M. Narayanan and M. Dawood, “Doppler
estimation using a coherent ultra-wideband
random noise radar,” IEEE Transactions on
Antennas & Propagation, 48(6): 868-878,
June 2000.
3. D.S. Garmatyuk and R.M. Narayanan,
“Ultrawideband noise synthetic aperture radar:
9
UNCLASSIFIED
(c) Theoretical output at 10 GHz frequency
(d) Same as above with expanded scale
theory and experiment,” Digest of the 1999
IEEE Antennas & Propagation Symposium,
Orlando, FL, 4: 2226-2229, July 1999.
4. D.C. Bell and R.M. Narayanan, “ISAR
imaging using a coherent ultra-wideband
random noise radar system,” Optical
Engineering, 40(11): 2612-2622, November
2001.
5. R.M. Narayanan, R.D. Mueller, and R.D.
Palmer, “Random noise radar interferometry,”
Proceedings of the SPIE Conference on Radar
Processing, Technology, and Applications,
Denver, CO, 2845: 75-82, August 1996.
6. D.S. Garmatyuk and R.M. Narayanan, “ECCM
capabilities of an ultra-wideband bandlimited
random noise imaging radar,” IEEE
Transactions on Aerospace & Electronic
Systems, 38: in press, 2002.